Google Clarifies That Disabled Source Clicks on Discover AI Summaries Were an Unintended Bug in an Experiment

By SEJ Staff | Updated October 2026

Google has officially addressed widespread concern within the search engine optimization (SEO) and digital publishing communities regarding a recent functionality change on Google Discover. According to Rajan Patel, Vice President of Engineering for Search at Google, an issue that temporarily prevented users from tapping or clicking on source citations within Google Discover’s artificial intelligence (AI) summary cards was the result of an unintended bug tied to a routine product experiment.

The glitch, which sparked immediate debate among SEO analysts and digital publishers reliant on Discover traffic, has since been resolved. Google confirmed that the specific experiment responsible for the blocked interaction has been shut down, restoring the original user experience.

1. Main Facts

The core of the issue centers on how Google Discover presents AI-generated summaries and the crucial links that direct users back to original content creators.

• The Incident: Users and researchers noticed that tapping on the source icon links displayed beneath AI-generated summaries on Google Discover failed to execute their intended function. Instead of leading readers to a feed of articles or the publisher’s original source material, the buttons were unresponsive.
• The Culprit: Google’s Rajan Patel confirmed on the social media platform X (formerly Twitter) that the disabled clicks were a bug stemming from an internal Discover feature experiment.
• Resolution: Google rapidly deactivated the problematic test. Patel verified that the functionality has been rolled back to how it was originally intended to work.
• Discovery: The issue was initially flagged and brought to light by search researcher Damien Andell, who monitors granular shifts in Google Discover’s interface and underlying code.

For publishers who depend on Google Discover as a significant source of referral traffic, any barrier between an AI summary card and the underlying source link represents a potential drop in readership and engagement. Consequently, Google’s prompt acknowledgment and swift remediation have provided a sigh of relief to the SEO community.

2. Chronology of Events

The unfolding of the Google Discover source-tap bug provides a clear timeline of how technical testing, community surveillance, and corporate transparency intersect in the modern search landscape.

October 23: The Discovery

Search researcher Damien Andell published findings on X indicating that Google Discover was testing a modification to its AI-generated summaries. Specifically, Andell uncovered an internal code flag labeled disable_aio_sources_button_tap. Testing revealed that this flag effectively blocked users from tapping on source attribution links located at the bottom of AI summary cards, rendering them unable to open the expected article feeds.

Later that same morning, prominent SEO expert Lily Ray quote-posted Andell’s findings, drawing broader industry attention to the potential implications for publisher traffic.

October 26–27: Official Response and Replacement Test

On Monday evening, October 26, Rajan Patel directly answered Lily Ray’s post on X. Patel clarified the situation, writing:

"Hey Lily — this was a bug in a Discover experiment. We’ve turned off the experiment, and the experience works as intended again."

The following morning, on October 27, Damien Andell conducted a follow-up investigation to verify Google’s claims. Confirming that the initial experiment had indeed been pulled, Andell noted that Google had immediately initiated a subsequent test. According to his findings, the previous flag was replaced by a new experiment titled "aio tap falls through to hero", signaling that Google continues to iterate on how users interact with AI summaries on mobile devices.

3. Supporting Data and Technical Analysis

While Google’s public-facing statements clarified that the disabled taps were an accident of experimentation, the technical mechanics uncovered by researchers highlight the complex nature of deploying generative AI features at scale.

Understanding the Experiment Flag

In modern software development, tech giants like Google rely heavily on feature flags—conditional configurations in code that allow engineers to roll out changes to a subset of users without deploying a full system update.

• The Flag: The disable_aio_sources_button_tap flag was ostensibly designed to test user behavior when source interaction was temporarily restricted.
• The Breakdown: According to visual evidence shared by Andell—which featured a YouTube video from Fox 9 appearing in a Discover feed alongside a brief AI summary and source icons—the interface displayed the visual elements of citation but stripped away the interactive capability.
• The Successor: The transition to the "aio tap falls through to hero" flag suggests that Google is actively testing alternative touch behaviors. While details on this new experiment remain sparse, it indicates that Google is experimenting with streamlining how clicks transition from the AI summary card to the primary piece of media or the "hero" content.

Google Search Console Metrics

The incident also shines a light on how visibility within generative AI features is tracked. The Google Search Console Discover Generative AI Report provides site owners with data regarding how often their URLs appear via Discover’s AI capabilities.

However, industry analysts note a distinct limitation in current metric tracking:

• Impressions vs. Clicks: The primary metric provided in Search Console for this report is the Impression—recording when a user visually encounters a website’s link via Discover’s generative AI.
• Attribution Nuances: Because the reporting focuses heavily on impressions, accurately measuring the exact click-through degradation caused by a bug like disable_aio_sources_button_tap can be challenging for individual publishers relying solely on out-of-the-box analytics.

4. Official Responses and Industry Reactions

The swift dialogue between independent search researchers, high-profile SEO consultants, and Google executives highlights a maturing communication channel between search engine developers and the optimization community.

Rajan Patel’s Clarification

As Google’s VP of Engineering for Search, Patel’s intervention was vital in quelling rumors that the tech giant was intentionally withholding traffic from publishers or testing a permanent model that stripped out outbound links. By explicitly classifying the issue as a "bug in a Discover experiment," Google framed the incident as a byproduct of routine, albeit flawed, internal product testing rather than a strategic policy shift.

The Publisher Perspective

For digital media companies and content creators, the episode underscores a persistent anxiety: the growing footprint of AI-generated summaries within discovery feeds. When AI features summarize content directly on the search or discovery page, publishers naturally worry about cannibalization—users consuming the summary and never visiting the source website.

When those source links are disabled—even temporarily due to a bug—it exacerbates these fears. Industry experts have continually stressed the necessity of robust attribution models to ensure that platforms utilizing generative AI continue to reward the original journalism and content creation that feeds these models.

5. Implications for SEO and Future Discover Updates

As Google continues to weave artificial intelligence deeper into Google Discover, Search, and other core surfaces, several key takeaways emerge for digital marketers, SEO professionals, and publishers:

1. The Price of Rapid Experimentation: Tech companies constantly run multivariate and A/B tests to optimize user engagement. However, as AI features dictate billions of content recommendations daily, even short-lived bugs can have measurable impacts on publisher metrics. Engineers must exercise heightened caution when testing core interactive elements like outbound links.
2. Monitoring UI Changes: The discovery by independent researchers like Damien Andell proves that monitoring code flags and interface shifts is more critical than ever. Publishers cannot rely solely on delayed analytics reports; understanding real-time interface modifications helps explain sudden traffic fluctuations.
3. The Evolving AI Interface: With Google replacing the bugged experiment with the "aio tap falls through to hero" test, it is clear that AI summary card architecture is still in a fluid, experimental state. Publishers should expect continuous UI adjustments as Google figures out the optimal balance between providing immediate AI answers and driving meaningful traffic to external websites.

What Publishers Should Do Next

In light of Google’s confirmation that the bug has been resolved, publishers do not need to take corrective technical actions on their sites. However, digital teams should actively monitor their Google Search Console Discover Generative AI reports for any unusual anomalies during the late October timeframe.

Furthermore, content creators are encouraged to manually test Google Discover feeds on their own mobile devices—tapping AI summary source cards to ensure that article feeds and outbound links continue to open smoothly. As Google refines its generative AI initiatives, keeping a close eye on user-experience touchpoints will remain a cornerstone of effective modern SEO strategy.